In modern electrical machine control systems, it is known to use a mathematical model of the machine as a basis for the control. These models, which comprise accurate mathematical equations and parameters of the machine, produce information concerning the electrical state of the machine for the control system. Many such models require accurate parameter values from the electrical machine. These parameters should be identified before use for every machine, since the values can differ greatly from machine to machine. One of these values that should be identified before or during the use of the machine is the value of stator resistance Rs.
The importance of the correctness of the estimation of the stator resistance arises from the fact, that the stator resistance is used in connection with the estimation of the stator flux. The stator flux is in turn used together with the rotor flux in various control topologies as controlled variables which should be as accurate as possible. The stator flux is usually calculated using a stator differential equation in the form ofψs=∫us−Rsisdt.  (1)
If the estimate for the stator resistance is not accurate, the calculation using integral drifts and causes erroneous results for the stator flux and rotor flux. This will in turn influence on the reliability of the whole control of the machine.
The stator resistance can be directly measured from stator windings before the start-up of the machine. There are also numerous methods for estimating the resistance during the use of the machine. Since the stator resistance is temperature dependent, the value of the resistance should be estimated quite often.
Document M. Aktas, H. I. Okumus: Neural network based stator resistance estimation in direct torque control of induction motor, IJCI Proceedings of Intl. XII Turkish Symposium on Artificial Intelligence and Neural Networks, Vol. 1, No: 1, July 2003, discloses a method for determining the stator resistance using active power balance equation. The balance equation is used in the document as an input to a resistance adaptation mechanism which is implemented as a neural network. The use of the active power produces results which are not satisfactory especially in the regenerating mode.
The problem in the known methods for determining the rotor flux is the inaccuracy which results mostly from inaccurate determination of the stator resistance and thus from erroneous stator and rotor fluxes.